MI

Devstral Small 2507 Rebased Vision

MistralGGUF by mradermacherUpdated 2025-12-25
Download on Hugging Face
Parameters23.6B
Context128K
Layers40
Architecturellama
Smallest quant0.0 GB
Quick answer

Devstral Small 2507 Rebased Vision needs about 16.2 GB of memory at Q4_K_M (a 14.3 GB file) with an 8K context window, so it fits on a 24 GB GPU. This repository has 24 quants, from 0.0 GB to 19.3 GB (Q6_K). Longer context needs more: at 32K tokens the KV cache alone adds about 5.4 GB.

Devstral Small 2507 Rebased Vision – 23.6B parameter vision model

Devstral Small 2507 Rebased Vision is a 23.6B parameter model based on Mistral architecture. It supports coding, tools, and vision tasks. The model was created by re‑weighting a Mistral‑Small base with Devstral weights and retains a 131072 token context window.

Good at

  • Coding assistance
  • Tool usage
  • Vision inference

Limits

  • No training or fine‑tuning after rebase
  • Vision support only via mmproj files
  • License is apache‑2.0

Best for

  • Generating code snippets
  • Responding to image prompts
  • Automating tool‑based workflows

Who it’s for. Developers and researchers needing a lightweight vision‑capable LLM. The model is suitable for local inference with llama.cpp and can be used in applications that require coding and image understanding.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
—0.0 GB 1.9 GB
Fits Get file →
IQ1_S5.3 GB 7.1 GB
Fits Get file →
IQ1_M5.8 GB 7.6 GB
Tight Get file →
IQ2_XXS6.5 GB 8.4 GB
Too big Get file →
IQ2_XS7.2 GB 9.0 GB
Too big Get file →
IQ2_S7.5 GB 9.3 GB
Too big Get file →
IQ2_M8.1 GB 10.0 GB
Too big Get file →
Q2_K_S8.3 GB 10.2 GB
Too big Get file →
Q2_K8.9 GB 10.7 GB
Too big Get file →
IQ3_XXS9.3 GB 11.1 GB
Too big Get file →
IQ3_XS9.9 GB 11.7 GB
Too big Get file →
Q3_K_S10.4 GB 12.2 GB
Too big Get file →
IQ3_S10.4 GB 12.3 GB
Too big Get file →
IQ3_M10.7 GB 12.5 GB
Too big Get file →
Q3_K_M11.5 GB 13.3 GB
Too big Get file →
Q3_K_L12.4 GB 14.2 GB
Too big Get file →
IQ4_XS12.8 GB 14.6 GB
Too big Get file →
Q4_013.5 GB 15.3 GB
Too big Get file →
Q4_K_S13.5 GB 15.4 GB
Too big Get file →
Q4_K_M14.3 GB 16.2 GB
Too big Get file →
Q4_114.9 GB 16.7 GB
Too big Get file →
Q5_K_S16.3 GB 18.1 GB
Too big Get file →
Q5_K_M16.8 GB 18.6 GB
Too big Get file →
Q6_K19.3 GB 21.2 GB
Too big Get file →

For 8 GB VRAM, use IQ1_S (5.27 GB, 7.1 GB memory at 8k context) or IQ1_M (5.75 GB, 7.6 GB). If 12 GB VRAM is available, Q2_K (8.89 GB, 10.7 GB) or Q3_K_S (10.4 GB, 12.2 GB) provide higher quality. Larger models such as Q4_K_M (14.33 GB, 16.2 GB) require 16 GB+ VRAM.

What GPU do you need for Devstral Small 2507 Rebased Vision?

These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.

  • 4 GB VRAM: the best quant that fits is the main file (0.0 GB file, about 1.9 GB in use).
  • 6 GB VRAM: the best quant that fits is the main file (0.0 GB file, about 1.9 GB in use).
  • 8 GB VRAM: the best quant that fits is IQ1_M (5.8 GB file, about 7.6 GB in use).
  • 12 GB VRAM: the best quant that fits is IQ3_XS (9.9 GB file, about 11.7 GB in use).
  • 16 GB VRAM: the best quant that fits is Q4_K_S (13.5 GB file, about 15.4 GB in use).
  • 24 GB VRAM: the best quant that fits is Q6_K (19.3 GB file, about 21.2 GB in use).

Which Devstral Small 2507 Rebased Vision quant should you choose?

Q4_K_M is the usual starting point for Devstral Small 2507 Rebased Vision: it balances file size and quality well. Here it is a 14.3 GB file.

If memory is tight, the smallest file is the first one in the table at 0.0 GB, with the quality trade-offs that come with fewer bits.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturellama
Parameters23.6B
Layers40
Trained context131,072 tokens
KV cache (fp16)160 KB per token

Source

Quantized bymradermacher
FamilyMistral
Licenseapache-2.0
Downloads (HF)1,168
Last updated2025-12-25

Frequently asked questions about Devstral Small 2507 Rebased Vision

How much VRAM does Devstral Small 2507 Rebased Vision need?

About 16.2 GB at Q4_K_M with an 8K context window. That is the 14.3 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.

Can I run Devstral Small 2507 Rebased Vision on an 8 GB GPU?

Yes, with IQ1_M: the 5.8 GB file needs about 7.6 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of Devstral Small 2507 Rebased Vision GGUF?

It depends on the quant: file is 0.0 GB, IQ1_S is 5.3 GB, IQ1_M is 5.8 GB, IQ2_XXS is 6.5 GB, IQ2_XS is 7.2 GB, IQ2_S is 7.5 GB and 18 more.

What context length does Devstral Small 2507 Rebased Vision support?

The GGUF header lists a trained context length of 131,072 tokens (128K). The KV cache grows with the context you use: about 1.3 GB at 8K and 5.4 GB at 32K.

Where can I download Devstral Small 2507 Rebased Vision GGUF?

The files are on Hugging Face in the mradermacher/Devstral-Small-2507-Rebased-Vision-i1-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

What license does Devstral Small 2507 Rebased Vision use?

Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.

Can I run Devstral Small 2507 Rebased Vision on 8 GB VRAM?

Yes, using IQ1_S or IQ1_M quantizations. They fit within 8 GB VRAM and use about 7–8 GB memory at 8k context.

Does the model support image input?

Yes, it is a vision model. Image handling requires mmproj files included in the static repository.

What license applies to this model?

The model is released under the apache‑2.0 license.

Other GGUFs of Devstral-Small-2507-Rebased-Vision

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Mistral models

All Mistral →

Other medium models

All medium models →